The inference speed of a computer vision model is directly proportional to the size of the input image. Moreover, dividing the dimensions of an image by two means four times fewer pixels for the model to process. Therefore, using smaller images improves inference speed.
When using smaller images, the model has less information and fewer details to work with. This often has an impact on the quality of the results. It is necessary to experiment with image size to find a good trade-off between speed and accuracy.
